Search Results

The RespArray™ patient monitor is intended to be used for monitoring, storing, of, and to generate alarms for, multiple physiological parameters of adults, pediatrics and neonates. The monitors are intended for use by trained healthcare professionals in hospital environments.
The monitor is for prescription use only.
The monitored physiological parameters include: ECG, respiration (RESP), temperature (TEMP), oxygen saturation of arterial blood (SpO2), pulse rate (PR), non-invasive blood pressure (NIBP), and carbon dioxide (CO2).
The arrhythmia detection and ST Segment analysis are intended for adult patients.
The SpO2 (NellcorTM) module is intended to be used for spot-check or continuous non-invasive monitoring of functional oxygen saturation of arterial hemoglobin (SpO2) and pulse rate (PR), in motion conditions, and in patients who are well or poorly perfused.
The MicrostreamTM capnography module is intended for continuous non-invasive monitoring of carbon dioxide concentration of the expired and inspired breath (etCO2) and respiration rate (RR).
The monitor also provides the clinician with integrated pulmonary index (IPI), apnea per hour (A/hr) and oxygen desaturation index (ODI) values. IPI is not intended for patients up to the age of one year. Allr and ODI are intended for ages 22 and up.
The monitors are not intended for MRI environments.

The RespArray patient monitor (hereinafter called RespArray) can perform long-time continuous monitoring of multiple physiological parameters. Also, it is capable of storing, displaying, analyzing and controlling measurements, and it will indicate alarms in case of abnormity so that doctors and nurses can deal with them in time.

The provided text is a 510(k) Premarket Notification summary for the "Patient Monitor: RespArray" device. This type of submission focuses on demonstrating substantial equivalence to legally marketed predicate devices, rather than conducting new clinical trials for de novo clearance or PMA approval. Therefore, the details requested about acceptance criteria, specific study design (like MRMC studies, sample sizes, expert ground truth establishment for AI/algorithm performance), and training set information are not typically found in these types of submissions, as the FDA review here centers on comparing the new device's specifications and performance to an existing, already cleared device.

The document primarily highlights the device's technical specifications and how they compare to a predicate device (Edan Instruments, Inc, Patient Monitor Model X8, X10, X12 - K192514), along with compliance with relevant electrical safety, EMC, and performance standards. It explicitly states "Clinical data: Not applicable."

Given this, I will extract the information that is present in the document and indicate where the requested information is not applicable or not provided within the scope of a 510(k) submission focused on substantial equivalence.

Analysis of the Provided Document for Device Acceptance Criteria and Study Proof

The provided document is a 510(k) premarket notification. For devices cleared via a 510(k), the primary "acceptance criterion" is often substantial equivalence to a legally marketed predicate device, demonstrated through comparative testing and adherence to recognized standards. Direct, explicit "acceptance criteria" presented as quantitative performance targets with a detailed study to prove they are met (as might be seen in AI/ML clearances for algorithms with novel functionalities) are typically not included in this type of submission for a patient monitor.

The "study" that proves the device meets the acceptance criteria is primarily non-clinical performance testing (bench testing) and software verification/validation to show that the device performs as intended and is as safe and effective as its predicate.

Here's a breakdown of the requested information based on the provided document:

1. A table of acceptance criteria and the reported device performance

The document does not provide a table of explicit acceptance criteria/performance targets with quantitative results in the way one might expect for a new AI/ML algorithm. Instead, it demonstrates performance by stating compliance with recognized consensus standards and by comparing the subject device's specifications to those of its predicate device, showing "similar design features and performance specifications."

The closest representation of "performance" and "acceptance" is the "Predicate Device Comparison" table (pages 5-6). This table implicitly acts as the performance comparison against the predicate device that serves as the "acceptance" benchmark for substantial equivalence.

The document concludes that "the subject and predicate devices have similar design features and performance specifications. The technological differences between the subject and predicate devices do not raise different questions of safety or effectiveness."

2. Sample size used for the test set and the data provenance

The document states "Clinical data: Not applicable." Therefore, there isn't a "test set" in the sense of patient data used for clinical validation of, for example, an AI algorithm's performance. The "testing" primarily refers to non-clinical bench testing.

• Sample size: Not applicable for patient data test set. For bench testing, samples would be physical devices, components, or simulated signals, but a "sample size" in terms of patient numbers is not provided.
• Data provenance (e.g., country of origin of the data, retrospective or prospective): Not applicable, as no clinical data test set was used/provided.

3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts

Not applicable. Since no clinical data test set was described and "Clinical data: Not applicable" is stated, there was no need for expert ground truth establishment for a test set. This type of information would be relevant for AI/ML device clearances where human expert annotation is part of the ground truth creation.

4. Adjudication method (e.g. 2+1, 3+1, none) for the test set

Not applicable, as no clinical test set requiring adjudication of ground truth was described.

5. If a multi reader multi case (MRMC) comparative effectiveness study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance

No. The document explicitly states "Clinical data: Not applicable." MRMC studies are typically for evaluating the impact of AI algorithms on human reader performance, which is not the scope of this 510(k) submission for a patient monitor.

6. If a standalone (i.e. algorithm only without human-in-the-loop performance) was done

The device is a patient monitor with various physiological parameter measurements and alarms. Its "performance" is inherent in its ability to accurately measure these parameters and detect events like arrhythmias. The non-clinical bench testing demonstrated its standalone performance by showing compliance with relevant standards (e.g., IEC 60601-2-27 for ECG, IEC 80601-2-30 for NIBP, ISO 80601-2-61 for pulse oximeter).

While not explicitly called "standalone algorithm performance" in the AI/ML sense, the "Performance testing-Bench" section (page 8) confirms that "Edan has conducted functional and system level testing to validate the performance of the results of the bench testing show that the subject device meets its accuracy specification and meet relevant consensus standards." This demonstrates the device's functional performance in isolation.

7. The type of ground truth used (expert consensus, pathology, outcomes data, etc.)

For the non-clinical bench testing, the "ground truth" would be established by:

• Reference instruments or calibrated signals (e.g., precise electrical signals for ECG, known pressure values for NIBP, calibrated gas mixtures for CO2).
• Standardized measurement protocols defined by the cited IEC/ISO standards.
• Accuracy specifications found within those standards or the device's own specifications.

There's no mention of expert consensus, pathology, or outcomes data as "ground truth" because this is a measurement and alarm device, not a diagnostic imaging AI algorithm, and no clinical data was used for validation in this submission.

8. The sample size for the training set

Not applicable. This device is a patient monitor, not an AI/ML algorithm that undergoes a distinct training phase on a dataset. The underlying algorithms for parameter measurement (e.g., NIBP oscillometric algorithm, arrhythmia detection) are established engineering designs, not typically "trained" in the machine learning sense with large datasets.

9. How the ground truth for the training set was established

Not applicable, as there is no specific "training set" for an AI/ML algorithm described. The "ground truth" for the development and calibration of the monitor's measurement algorithms would have been established through engineering principles, laboratory testing with calibrated instruments, and referencing physiological models and data, but this is part of the device's fundamental design and not a separate "training set" as understood in current AI/ML contexts.

Ask a specific question about this device